Since there are no reliable and specific tests for suicidal behav-ior, the clinician must rely on clinical, demographic, historical and patient self-report information to guide their judgment and to tailor their intervention. However, the clinician should not be lulled into a false sense of security because of their assessment. The positive predictive value of individual or group of risk fac-tors is low. Stepwise multiple regression was used to develop a statistical model that would predict suicide in 1906 inpatients (admitted between 1970 and 1981). There were 46 suicides at the end of the follow-up period (1983). Medical illness, substance abuse, personality disorder, marital status were among charac-teristics that failed to reach significance for the final model. Vari-ables that met significance level criteria (number of prior suicide attempts, suicidal ideation, bipolar disorder, gender, outcome at discharge, unipolar depression with bipolar first-degree relative) were included in the model. The model did not identify any pa-tient who committed suicide. Limitations include the low suicide rate and possible underestimation of suicides (Goldstein et al., 1991). There may be differences in people who attempt suicide versus those who complete it. In 1999, it was estimated that there were 730 000 annual attempts, with an attempt and completion ratio of 25 : 1. It was estimated that 5 million had a history of a sui-cide attempt (Hoyert et al., 2001). Differential risk factors include gender, ethnicity, age, substance abuse, past suicide attempts, means, family history, hopelessness, comorbid medical condi-tions, relationship/esteem losses resulting in murder–suicide and media coverage (see Table 85.2).